Chemometric Strategies for Simultaneous Quantification of Co-Existing Pollutants in Competitive and Multicomponent Adsorption Systems: A Critical Review

Abstract

Adsorption studies are still, in the great majority of cases, reported for a single adsorbate, yet the effluents they claim to treat almost never contain one pollutant. Moving from one solute to two or three changes the experiment in a way that is easy to underestimate, because the analytical step itself becomes the limiting factor: the absorbance recorded at one wavelength can no longer be assigned to one species. This review examines how chemometric tools have been used, and occasionally misused, to recover individual concentrations of co-existing pollutants during competitive adsorption. We first set out the artefacts peculiar to adsorption supernatants: spectral overlap, adsorbent-derived colour, residual turbidity and pH-driven speciation. We then follow the methodological ladder from univariate remedies such as derivative and ratio-spectra techniques and the H-point standard addition method, through first-order multivariate calibration by principal component regression and partial least squares with wavelength selection and net analyte signal theory, up to second-order and multiway approaches such as curve resolution and parallel factor analysis, which retain accuracy in the presence of uncalibrated interferents. Neural networks and machine-learning regressors are discussed separately, since they address a genuinely nonlinear problem but carry their own validation burden. Throughout we insist on a point that most application papers pass over: competition factors, extended Langmuir constants and ideal adsorbed solution predictions inherit the uncertainty of the quantification step, so an unvalidated calibration propagates quietly into the thermodynamic conclusions. A minimum reporting set is proposed to make multicomponent removal data comparable between laboratories.